Memetic Algorithms for the Workforce Routing and Scheduling Problem: case study of Energy Distribution System Operator (DSO)
Résumé
This study addresses a real industrial problem faced by a company operating an
energy distribution network: the efficient assignment of tasks to their agents. This problem involves determining the best routes for agents to visit multiple tasks, known in the literature as the Technician Routing and Scheduling Problem (TRSP). It considers a limited number of agents, and tasks located in different areas. Each task has a service duration, a time window during which the service should be executed, and a set of required skills. Additionally, agents
possess different skills, multiple availability intervals, and a maximum daily working time. The objective is to minimize the total traveled distance, including penalties for unrouted tasks. To tackle this problem, we propose an order-first split-second approach that combines a memetic algorithm using a giant tour encoding of the chromosome with an extension of the optimal split method. Our primary objective is to evaluate the effectiveness of this approach on a real case of the TRSP problem in the context of energy network management. The second objective is to compare our genetic algorithm, which uses the giant tour encoding with a genetic algorithm scheme that employs an indirect encoding representation proposed in the literature. It is worth noting that this indirect encoding has proved its effectiveness on real instances of an industrial problem, which is similar to our case study. We aim to test different implementations of the proposed genetic algorithms on real instances to evaluate the impact of the extended optimal split, local search, and encoding type.